Not every task needs a GPU, but the list of development scenarios that benefit from GPU resources is expanding\u2014from experimenting with AI models and running inferencing tasks to processing large datasets or doing intensive simulations. Until now, when a developer hit a point in the project when they needed a GPU, they would have to go through a slow process to configure Azure resources and get permissions, and supporting admins would have to manage the GPU infrastructure as well. This approach adds unnecessary toil for everyone involved and can be prohibitively expensive. Now, Dev Box simplifies access to serverless GPU resources and avoids the added costs from dedicated GPUs. In a couple steps, platform engineers can add serverless GPU to the projects that need it, and you can get the compute power you need, when you need it.<\/p>\n
But not every AI task requires a GPU. For scenarios where you need to get started with existing AI models, we are excited to announce that you can now access preconfigured Azure AI resources<\/strong><\/a> directly from your dev boxes (Preview). Like the Dev Box serverless GPU experience, platform engineers can easily configure Azure AI resources for specific development projects so that you can get a zero-config experience when accessing large language models (LLMs) from Azure AI Foundry.<\/p>\n Direct access to LLMs in the development environment streamlines key tasks like prompt tuning, model comparison, and building agent-based workflows. When building an AI app, this means faster iteration, the ability to test ideas in real time, and minimum setup time when integrating powerful LLM capabilities into your app.<\/p>\n Beyond advancing AI development, Dev Box now further elevates developer productivity through the new Dev Box<\/strong> Model Context Protocol (MCP) server<\/strong><\/a> (Preview). This provides developers with a faster, more intelligent way to interact with Dev Box infrastructure from within their Visual Studio Code environment to help eliminate context switching and speed up workflows.<\/p>\n Now you can use conversational prompts to create, customize, and control dev boxes without ever leaving your IDE. Instead of going to the Dev Box developer portal, you can simply ask GitHub Copilot agent mode to do things like give you a list of your running dev boxes, create a new one, or perform actions like shutdown, restore, troubleshoot, delete, and more\u2014all from the chat window in VS Code\u2014for a more intuitive, focused development experience. For example, imagine you want to experiment with your code but are worried it might break your setup. You can now take a manual snapshot of your dev box<\/strong><\/a> (Preview), and if something goes wrong, you can revert to a previous working state with a natural language prompt in your IDE. This way, you can experiment with confidence, knowing that you can easily maintain your current setup. And to further streamline your workflow and minimize disruptions, you can now open your dev box<\/strong> directly from the VS Code<\/strong> extension<\/strong><\/a> (Preview) \u2014ideal for scenarios that don\u2019t require an RDP connection or involve high-latency networks.<\/p>\n Project onboarding is often a major hurdle in software development, especially for enterprise teams working on complex, AI-powered applications that require extensive tooling, large dependencies, and specialized environments. Configuration challenges slow down teams before they even start coding. Dev Box solves this problem by enabling dev leads to quickly create and scale custom configurations for their teams regardless of project complexity.<\/p>\n To further simplify and scale project-based environments, we are excited to announce the general availability of team customizations<\/strong><\/a> and Dev Box imaging<\/strong><\/a>. With team customizations, dev leads can define a single configuration file with all the necessary tools, packages, and settings required for a project, while developers get ready-to-code environments at the click of button, eliminating configuration chores and ensuring consistent experiences across the team.<\/p>\n![\"Crop](\"https:\/\/devblogs.microsoft.com\/develop-from-the-cloud\/wp-content\/uploads\/sites\/81\/2025\/05\/Crop-Serverless-GPU-1024x334.png\")
<\/a><\/p>\nCreate custom dev environments for teams with an agentic workflow<\/h3>\n